AI in Payment Integrity: Why the Foundation Matters More Than the Algorithm

Payment integrity leaders are operating in one of the most challenging payer environments in over a decade. Margins are tightening. Utilization is rising, driven by specialty drugs and high-acuity care. Administrative costs are climbing. And provider appeals are becoming faster and more sophisticated – often powered by AI themselves.

Against this backdrop, AI has become a paradox. The same force amplifying the pressure is now the only viable response to it.

According to research from McKinsey & Company, roughly 80% of organizations have deployed generative AI in some form. Yet most report ambiguous results. Seen as a catalyst for innovation, AI use is broadening but is still in pilot stages. In healthcare, that gap between experimentation and enterprise value is especially pronounced. Those realizing the largest returns from AI are using best practices, which include ensuring AI is part of a broader closed loop system vs. used as a point solution.

“The impact of AI in payment integrity is driven by workflow integration, not model sophistication,” said Udit Anand, associate partner at McKinsey & Company, in a recent executive discussion on AI and payment integrity.

That insight gets to the heart of the problem.

The AI Experiment Trap

AI isn’t a quick fix. Payment integrity data fragmentation persists. And, as Anand pointed out, many health plans are layering increasingly sophisticated models onto legacy systems that were never designed to work together. Prepay edits, post-pay recovery data, and appeals tracking sit in different systems and spreadsheets. Upstream of payment integrity, policy logic and contract interpretation may not be standardized or version-controlled across systems.

In this environment, AI agents may still summarize documentation, calculate DRGs, or draft audit rationales. But when the underlying data is incomplete, inconsistent, or siloed, AI can’t do what it does best – learn from outcomes that are then fed back into the model.

The result is familiar: recurring errors, rising overturn rates, and increasing administrative burden.

AI isn’t a strategy in and of itself. AI needs to be built upon and integrated within a strong data foundation to progress beyond the experiment stage.

The Case for a Closed-Loop Foundation

As AI adoption realities emerge, health plans are re-thinking their deployment strategies. Leading payment integrity functions are shifting from point solutions to closed-loop data environments that connect the full claim lifecycle – from pre-adjudication through post-pay and appeals.

In practice, that means capturing payment accuracy activity in a single, unified system where key decisions and rationale are preserved at every stage. Post-pay findings reveal what slipped through to refine prepay rules. Appeals outcomes become structured inputs into future editing logic and policy updates.

Connecting all payment integrity activity creates a feedback loop – key when turning AI from a pilot into a performance engine.

This architecture also enables more advanced workflows, including coordinated AI agents that retrieve documentation, cross-check policy logic, validate contracts, and flag exceptions for human review. And it can more easily bring a human expert into the review stream, and more readily capture rationale and explainability.

In regulated environments such as payment integrity, “human in the loop” oversight is essential. But as Tom Noack, long-time payment integrity leader and SVP of product strategy for ClarisHealth, notes, the human role shifts. This evolution from manual processing to judgment, exception management, and governance unlocks the scale payers need.

“As payers contend with high medical and administrative costs, one platform driving insights and data and taking advantage of integrated AI is the key to payment integrity operational success,” says Noack.

Why This Matters Now

Providers are moving fast. Ambient documentation and AI-assisted coding are already reshaping what lands in a payer’s queue — shifting coding intensity and accelerating appeal sophistication. But investment isn’t translating to results. A joint report from Bessemer Venture Partners, Bain & Company, and AWS found that while more than 80% of healthcare leaders believe AI will transform clinical decision-making, only 30% of pilots reach production.

Payer organizations that build a unified data layer across claims, utilization management, network and payment integrity don’t just reduce complexity — they create a structural advantage built for scale:

• • They gain visibility into patterns that siloed systems obscure.
  • They can measure balanced metrics – recovery, denial rates, overturn rates and provider abrasion – across the lifecycle.
  • They can embed AI directly into workflows rather than bolting it on.

The Strategic Role of Payment Integrity Leaders

Not every AI decision sits within the payment integrity function. Enterprise strategy, technology architecture, and capital allocation often reside at the C-suite level. But the data lives closer to home.

Payment integrity leaders who own the closed-loop data foundation don’t just inform enterprise AI strategy – they help shape it.

They know where the signal is. They surface data gaps. They advocate for governance clarity.  And they know which AI investments will compound versus which ones will stall.

Organizations that rewire their data and workflows before layering in AI will scale innovation, reduce administrative burden and improve provider experience. Those that don’t will amplify the inefficiency they were trying to solve.

Originally published in Fierce Health Payer